Phrony
Phrony (Developer: Phrony Platform, release date: July 2026) is an open-source runtime and governance framework that enables the deployment and management of AI agents as independent, first-class primitives based on declarative YAML manifests. Phrony separates the operational principles and lifecycle of agents, which were previously managed within complex Python scripts, into a single, unified configuration file. Similar to how container virtualization technology like Docker virtualizes web servers in isolated environments.
Phrony (Developer: Phrony Platform, scheduled for release in July 2026) is an open-source runtime and governance framework that enables the deployment and management of AI agents as independent, first-class primitives based on declarative YAML manifests. Phrony separates the operational principles and lifecycle of agents, which were previously managed within complex Python scripts, into a single, unified configuration file. Similar to how Docker, a container virtualization technology, reliably runs web servers in isolated environments, Phrony Runtime operates by completely isolating LLM inference loops, tool invocation mediation, cost control, and policy rules into independent sessions through a central gRPC daemon architecture.
Traditional agent development involved tightly coupling libraries like LangChain or CrewAI at the application source code level, which limited the ability to implement complex error handling and loop control. This resulted in agents being prone to unexpected malfunctions, leading to infinite loops, excessive API costs, or uncontrolled execution of unverified and potentially dangerous tools. Phrony declaratively controls the agent's lifecycle, similar to how Kubernetes schedules containers using YAML definitions. It also provides a safe autonomous operation environment by declaring human-in-the-loop trigger settings in the manifest, ensuring that the agent's execution is paused and awaits human approval before performing critical actions.
Introducing Phrony into AI workflows for life sciences and pharmaceutical development allows for highly precise and secure orchestration of distributed research analysis resources and genomic data processing pipelines. For example, when a large-scale variant interpretation agent collects information from the Ensembl API or ClinVar database and predicts the pathogenicity of variants, it monitors pre-declared token limits and malfunction restriction thresholds to prevent runaway costs. Furthermore, in the final statistical validation of high-risk variants or the external transmission of sensitive information, the agent's actions are halted, and notifications are sent to human researchers to obtain approval, thereby maximizing the security and reliability of complex and sensitive research and development.
💻 System Requirements
0 (For CPU only. However, when running sessions via a local LLM, follow the hardware specifications required to run that local LLM.)
500MB and below for binary and basic dependency package installations (lightweight when using Docker Compose's default configuration)
⚡ Installation
4-1. Quick Start
go install github.com/phrony-platform/runtime/cmd/cli@latest
mv -f "$(go env GOPATH)/bin/cli" "$(go env GOPATH)/bin/phrony"
4-2. Detailed installation
# 1. Create and navigate to the working directory
mkdir phrony-quickstart && cd phrony-quickstart
# 2. Download Docker Compose file
curl -fsSLO https://phrony.com/runtime/docker-compose.yml
# 3. Running PostgreSQL and Phony runtime daemon in the background
docker compose up -d --wait
# 4. Check runtime connection status
phrony status
🧬 Bio Use Cases
Automated Curation and Annotation of Cancer Genomic Variants
Deploy a Phrony agent that integrates the ClinVar API and the Ensembl VEP tool to annotate 1,000 variant data points within 5 minutes. When high-risk variants (CADD > 20) are detected, trigger a HITL approval gate to support real-time clinical decision-making.
Automated Large-Scale Virtual Screening and Toxicity Filtering
Declare an agent that combines the ChEMBL database API and the AutoDock Vina tool using a YAML manifest. This agent screens 200 lead compounds with a ligand binding energy of -8.0 kcal/mol or less, reducing costs by 20% and obtaining final researcher approval.
Automated Clinical Trial Protocol Compliance Verification and Reporting
Utilize an agent that connects the ClinicalTrials API and a local Llama-3-70B model to automatically analyze 50 target disease clinical design data sets. By blocking duplicate criteria under the constraint of max_loop_iterations: 10, it automatically generates a precise report.
FAQ
What is Phrony?
Phrony (Developer: Phrony Platform, scheduled for release in July 2026) is an open-source runtime and governance framework that enables the deployment and management of AI agents as independent, first-class primitives based on declarative YAML manifests. Phrony separates the operational principles and lifecycle of agents, which were previously managed within complex Python scripts, into a single, unified configuration file. Similar to how Docker, a container virtualization technology, reliably runs web servers in isolated environments, Phrony Runtime operates by completely isolating LLM inference loops, tool invocation mediation, cost control, and policy rules into independent sessions through a central gRPC daemon architecture. Traditional agent development involved tightly coupling libraries like LangChain or CrewAI at the application source code level, which limited the ability to implement complex error handling and loop control. This resulted in agents being prone to unexpected malfunctions, leading to infinite loops, excessive API costs, or uncontrolled execution of unverified and potentially dangerous tools. Phrony declaratively controls the agent's lifecycle, similar to how Kubernetes schedules containers using YAML definitions. It also provides a safe autonomous operation environment by declaring human-in-the-loop trigger settings in the manifest, ensuring that the agent's execution is paused and awaits human approval before performing critical actions. Introducing Phrony into AI workflows for life sciences and pharmaceutical development allows for highly precise and secure orchestration of distributed research analysis resources and genomic data processing pipelines. For example, when a large-scale variant interpretation agent collects information from the Ensembl API or ClinVar database and predicts the pathogenicity of variants, it monitors pre-declared token limits and malfunction restriction thresholds to prevent runaway costs. Furthermore, in the final statistical validation of high-risk variants or the external transmission of sensitive information, the agent's actions are halted, and notifications are sent to human researchers to obtain approval, thereby maximizing the security and reliability of complex and sensitive research and development.
When should I use Phrony?
Phrony (Developer: Phrony Platform, release date: July 2026) is an open-source runtime and governance framework that enables the deployment and management of AI agents as independent, first-class primitives based on declarative YAML manifests. Phrony separates the operational principles and lifecycle of agents, which were previously managed within complex Python scripts, into a single, unified configuration file. Similar to how container virtualization technology like Docker virtualizes web servers in isolated environments.
What is a biomedical use case for Phrony?
Automated Curation and Annotation of Cancer Genomic Variants: Deploy a Phrony agent that integrates the ClinVar API and the Ensembl VEP tool to annotate 1,000 variant data points within 5 minutes. When high-risk variants (CADD > 20) are detected, trigger a HITL approval gate to support real-time clinical decision-making.
📝 Update Notes
- v0.2.09/14/2026
Phrony 0.2.0 업데이트를 통해 이제 OpenAI와 호환되는 다양한 AI 모델 제공자를 자유롭게 연결할 수 있게 되었어요. 이를 통해 민감한 생물학적 데이터를 다룰 때 보안을 위해 로컬 모델을 활용하거나, 특정 바이오 데이터에 특화된 맞춤형 LLM을 연구 프로세스에 도입하기가 훨씬 수월해졌습니다. 또한 CLI 업그레이드 기능과 버그 수정이 포함되어 연구 도구의 관리와 안정성도 한층 높아졌습니다.
- v0.1.18/8/2026
Phrony 0.1.1 업데이트에서는 번들을 배포할 때 특정 버전의 외부 에이전트를 참조하면 발생하던 오류가 수정되었어요. 이제 분석 파이프라인에 꼭 필요한 특정 버전의 도구를 안정적으로 고정하여 배포할 수 있게 되었습니다. 실험 결과의 재현성이 핵심인 바이오 연구에서, 도구 버전 관리를 더욱 정확하게 수행하여 워크플로우의 신뢰도를 높이는 데 큰 도움이 될 거예요.
- v0.0.27/6/2026
Full Changelog: https://github.com/phrony-platform/runtime/compare/0.0.1...0.0.2
🧪 Related Code of Life
No related Code of Life posts yet.